TWI793653B - Information processing device and information processing method - Google Patents

Information processing device and information processing method Download PDF

Info

Publication number
TWI793653B
TWI793653B TW110123302A TW110123302A TWI793653B TW I793653 B TWI793653 B TW I793653B TW 110123302 A TW110123302 A TW 110123302A TW 110123302 A TW110123302 A TW 110123302A TW I793653 B TWI793653 B TW I793653B
Authority
TW
Taiwan
Prior art keywords
image
images
state
item
aforementioned
Prior art date
Application number
TW110123302A
Other languages
Chinese (zh)
Other versions
TW202217705A (en
Inventor
中澤満
*** 艾奎瑪
友岡高志
Original Assignee
日商樂天集團股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日商樂天集團股份有限公司 filed Critical 日商樂天集團股份有限公司
Publication of TW202217705A publication Critical patent/TW202217705A/en
Application granted granted Critical
Publication of TWI793653B publication Critical patent/TWI793653B/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20076—Probabilistic image processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30168—Image quality inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

[課題] 從影像而更正確地判定出對象物為被特定之物品之狀態。 [解決手段] 資訊處理裝置係取得關於特定之物品的複數個影像，針對前記複數個影像之每一者而決定對於前記物品之狀態之推定的適合性，基於前記決定之結果、與前記複數個影像之至少一部分，而將表示前記特定之物品的已被推定之狀態的狀態資訊予以輸出。 [Problem] More accurately determine the state of the object as a specified item from the image. [Solution] The information processing device obtains a plurality of images of a specific item, and determines the suitability of the estimation of the state of the aforementioned item for each of the aforementioned plurality of images, based on the results of the aforementioned determination and the results of the aforementioned plurality of images. At least a part of the image is used to output the state information indicating the estimated state of the item specified above.

Description

資訊處理裝置及資訊處理方法Information processing device and information processing method

本發明係有關於資訊處理裝置及資訊處理方法。The present invention relates to an information processing device and an information processing method.

在中古商品等之對象物為被特定之物品的處理時，為了使該物品適切地流通，掌握其狀態是很重要的。例如在線上的跳蚤市場服務或競標中，商品的影像及狀態之排名會被公開。購入側的使用者係根據該排名來推定該商品的狀態。為了更容易掌握物品之狀態，自動判定物品之狀態的技術係已被開發。When dealing with specified items such as second-hand goods, it is important to grasp the state of the item in order to distribute the item appropriately. For example, in online flea market services or bidding, images of products and status rankings will be disclosed. The user on the purchase side estimates the status of the product based on the ranking. In order to make it easier to grasp the state of the item, the technology to automatically determine the state of the item has been developed.

日本特開2019-91323號公報中係揭露，基於商品影像來判定商品之狀態。Japanese Patent Application Laid-Open No. 2019-91323 discloses that the state of the product is determined based on the image of the product.

根據為了對象物為被特定之物品之說明而被登錄的商品影像，有時候難以自動判定該物品之狀態。例如，為了中古商品之說明而被登錄的影像，含有由製造者所提供之新品的影像的情況下，則會難以根據這些影像來適切地判定商品之狀態。It may be difficult to automatically determine the status of the item based on the product image registered for the description that the object is a specified item. For example, if images registered for explaining second-hand goods include images of new products provided by the manufacturer, it may be difficult to appropriately determine the state of the product based on these images.

本發明係有鑑於上記課題而研發，其目的在於，提供可根據影像而更正確地判定對象物為被特定之物品之狀態的技術技術。The present invention has been developed in view of the above-mentioned problems, and an object thereof is to provide a technology capable of more accurately judging that a target object is a specified article based on an image.

為了解決上記課題，本發明所述之資訊處理裝置係含有：取得部，係取得關於特定之物品的複數個影像；和決定部，係針對前記複數個影像之每一者而決定對於前記物品之狀態之推定的適合性；和輸出部，係基於前記決定部之決定結果、與前記複數個影像之至少一部分，而將表示前記特定之物品的已被推定之狀態的狀態資訊，予以輸出。In order to solve the above-mentioned problems, the information processing device of the present invention includes: an acquisition unit that acquires a plurality of images of a specific item; Estimated suitability of the state; and the output unit outputs state information representing the estimated state of the aforementioned specific item based on the decision result of the aforementioned determining unit and at least a part of the aforementioned plurality of images.

又，本發明所述之資訊處理方法係含有：取得關於特定之物品的複數個影像之步驟；和針對前記複數個影像之每一者而決定對於前記物品之狀態之推定的適合性之步驟；和基於前記決定之結果、與前記複數個影像之至少一部分，而將表示前記特定之物品的已被推定之狀態的狀態資訊予以輸出之步驟。In addition, the information processing method of the present invention includes: a step of obtaining a plurality of images of a specific item; and a step of determining the suitability of the presumption of the state of the aforementioned item for each of the aforementioned plurality of images; and a step of outputting state information representing an estimated state of the aforementioned specified item based on the result of the aforementioned decision and at least a part of the aforementioned plurality of images.

在本發明的一形態中，前記決定部，係可將前記複數個影像之每一者的權重，加以決定；前記輸出部，係可基於前記複數個影像之每一者，而將表示前記特定之物品之狀態的狀態要素值加以決定，並基於針對前記複數個影像之每一者而被決定的狀態要素值及前記權重，而生成表示前記特定之物品之狀態的狀態資訊，並將前記已被生成之狀態資訊予以輸出。 In one aspect of the present invention, the prescriptive determination unit can determine the weight of each of the plural images in the preface; The value of the state element of the state of the object is determined, and based on the value of the state element determined for each of the plurality of images in the foregoing and the weight in the foregoing, the state information indicating the state of the specified object in the foregoing is generated, and the foregoing has been The generated status information is output.

在本發明的一形態中，前記決定部，係可以使得前記取得部所取得之影像係為拍攝了前記特定之物品之影像的或然性越高，該當影像的權重就會越大的方式，來決定權重。 In one aspect of the present invention, the preamble determining unit can make the image acquired by the preamble obtaining unit higher in probability that the image of the item specified in the aforesaid is taken, the greater the weight of the corresponding image, to determine the weight.

在本發明的一形態中，前記決定部係可含有：藉由表示影像與前記影像之適合性的教師資料而被進行學習的已學習模型。 In one aspect of the present invention, the antecedent determination unit may include a learned model that is learned using teacher data indicating the compatibility between the video and the antecedent image.

在本發明的一形態中，前記決定部，係可檢索與前記已被取得之複數個影像相同之影像，基於相同之影像是否藉由檢索而被找到，來決定前記適合性。 In one aspect of the present invention, the affidavit determination unit can search for images identical to the plurality of images obtained in the affidavit, and determine affiliation suitability based on whether or not the same image is found through the search.

在本發明的一形態中，前記決定部，係可檢索與前記已被取得之複數個影像相同之影像，基於相同之影像是否藉由檢索而被找到，來決定前記權重。 In one aspect of the present invention, the affidavit determining unit can search for images identical to a plurality of images obtained in the affidavit, and determine the affiliation weight based on whether or not the same image is found through the search.

在本發明的一形態中，前記決定部，係可在前記相同之影像是藉由檢索而有被找到的情況下，基於前記已被找到之影像的提供來源，來決定權重。 In one aspect of the present invention, the affidavit determination unit may determine the weight based on the source of the image whose affiliation has been found, when an image with the same affiliation is found by searching.

在本發明的一形態中，前記決定部，係可將前記取得部所取得之影像係為前記特定之物品所被拍攝之影像的或然性之高低，當作前記適合性而加以決定。 In one aspect of the present invention, the affidavit determining unit can determine whether the image acquired by the affidavit acquiring unit is an image captured by the affidavit-specified article as the affidavit suitability.

藉由本發明，可容易從影像來判定對象物為 被特定之物品之狀態。 With the present invention, it is easy to judge the object from the image as The state of the specified item.

以下，基於圖式來說明本發明的實施形態。對標示相同符號的構成，係省略重複的說明。在本實施形態中係說明，一旦藉由使用者針對對象物已被特定之物品而登錄了含有影像的資訊，就將該已被登錄之資訊讓購入候補者進行瀏覽的資訊處理系統。資訊處理系統，係為例如作為對象物已被特定之物品而販售中古商品的中古商品之販售系統。 Hereinafter, embodiments of the present invention will be described based on the drawings. Duplicate explanations are omitted for components marked with the same symbols. This embodiment describes an information processing system in which once a user registers information including an image for an object whose object has been specified, the registered information is allowed to be browsed by purchase candidates. The information processing system is, for example, a sales system for second-hand goods that sells second-hand goods as objects whose objects have been specified.

圖1係為本發明的實施形態所述之資訊處理系統之一例的圖示。資訊處理系統係含有資訊處理伺服器1和1或複數個顧客終端2。顧客終端2，係為例如智慧型手機或個人電腦等，係被資訊處理系統所提供之服務的使用者進行操作。FIG. 1 is a diagram showing an example of an information processing system according to an embodiment of the present invention. The information processing system includes information processing servers 1 and 1 or a plurality of customer terminals 2 . The customer terminal 2 is operated by the user of the service provided by the information processing system, such as a smartphone or a personal computer.

資訊處理伺服器1，係與1或複數個顧客終端2進行通訊，從使用者所操作的顧客終端2接收關於對象物為被特定之物品的說明資訊與複數個影像，將說明資訊及影像登錄至服務內。又，對身為購入候補者的使用者，將關於物品的說明資訊及影像予以提示。對象物為被特定之物品，係為例如中古商品，即使在有同種類的複數個物品的情況下，購入者仍只會特定出其中任一物品而予以購入的商品。對象物為被特定之物品，係亦記載成「特定之物品」。以下為了簡化記載，在沒有特別聲明的情況下，「物品」係指對象物為被特定之物品(特定之物品)。The information processing server 1 communicates with one or a plurality of customer terminals 2, receives explanatory information and a plurality of images about the specified object from the customer terminal 2 operated by the user, and registers the explanatory information and images to the service. In addition, explanatory information and images of items are presented to users who are purchase candidates. The target object is a specified article, such as a second-hand commodity. Even if there are a plurality of articles of the same type, the purchaser can only specify any one of the articles and purchase it. The object is a specified item, and it is also described as a "specified item". In order to simplify the description below, unless otherwise stated, "article" refers to a specified article (specified article).

資訊處理伺服器1係含有：處理器11、記憶部12、通訊部13、輸出入部14。此外，資訊處理伺服器1，係為伺服器電腦。資訊處理伺服器1之處理，係亦可藉由複數個伺服器電腦來加以實現。The information processing server 1 includes: a processor 11 , a storage unit 12 , a communication unit 13 , and an input/output unit 14 . In addition, the information processing server 1 is a server computer. The processing of the information processing server 1 can also be realized by a plurality of server computers.

處理器11，係依照記憶部12中所儲存的程式而動作。又，處理器11控制通訊部13、輸出入部14。此外，上記程式係可透過網際網路等來提供，也可儲存在快閃記憶體或DVD-ROM等之電腦可讀取之記憶媒體中來提供。The processor 11 operates according to a program stored in the memory unit 12 . Furthermore, the processor 11 controls the communication unit 13 and the input/output unit 14 . In addition, the above-mentioned program may be provided through the Internet or the like, and may be stored in a computer-readable storage medium such as a flash memory or DVD-ROM.

記憶部12係由RAM及快閃記憶體等之記憶體元件與硬碟機這類外部記憶裝置等所構成。記憶部12係儲存上記程式。又，記憶部12，係將從處理器11、通訊部13、輸出入部14所被輸入之資訊或演算結果，加以儲存。記憶部12，係可將已被接收之說明資訊和複數個影像，加以儲存。又，為了儲存已被接收之說明資訊和複數個影像，亦可設置異於資訊處理伺服器1的儲存區。The memory unit 12 is composed of memory elements such as RAM and flash memory, and external memory devices such as hard disk drives. The memory unit 12 stores the above-mentioned program. Moreover, the memory unit 12 stores information or calculation results input from the processor 11, the communication unit 13, and the input/output unit 14. The memory unit 12 can store the received explanatory information and multiple images. Also, a storage area different from the information processing server 1 may be provided in order to store the received explanatory information and a plurality of images.

通訊部13，係實現與其他裝置進行通訊之機能，是由例如將無線LAN、有線LAN加以實現的積體電路等所構成。通訊部13，係基於處理器11之控制，而將從其他裝置收到的資訊，輸入至處理器11或記憶部12，並將資訊發送至其他裝置。The communication part 13 realizes the function of communicating with other devices, and is constituted by, for example, an integrated circuit realizing wireless LAN and wired LAN. The communication part 13, based on the control of the processor 11, inputs the information received from other devices into the processor 11 or the memory part 12, and sends the information to other devices.

輸出入部14，係由：控制顯示輸出裝置的視訊控制器、或從輸入裝置取得資料的控制器等所構成。作為輸入裝置係有鍵盤、滑鼠、觸控面板等。輸出入部14，係基於處理器11之控制，而對顯示輸出裝置輸出顯示資料，並將使用者藉由操作輸入裝置所輸入之資料加以取得。顯示輸出裝置係為例如被連接在外部的顯示器裝置。The input/output unit 14 is composed of a video controller for controlling the display output device, or a controller for obtaining data from the input device, and the like. As the input device, there are a keyboard, a mouse, a touch panel, and the like. The input/output unit 14 is based on the control of the processor 11 to output display data to the display output device, and obtain the data input by the user through the operation of the input device. The display output device is, for example, an externally connected display device.

接著，說明資訊處理系統所提供之機能。圖2係為資訊處理系統所實現之機能的區塊圖。資訊處理系統係含有：物品資訊取得部51、適合決定部52、狀態輸出部53。又，狀態輸出部53係在機能上，含有影像推定部54、狀態算出部55。這些機能，係藉由資訊處理伺服器1中所包含的處理器11來執行記憶部12中所被儲存的程式，並控制通訊部13等來加以實現。Next, the functions provided by the information processing system will be explained. FIG. 2 is a block diagram of the functions realized by the information processing system. The information processing system includes: an article information acquisition unit 51 , a suitability determination unit 52 , and a status output unit 53 . Furthermore, the status output unit 53 is functionally including a video estimation unit 54 and a status calculation unit 55 . These functions are realized by the processor 11 included in the information processing server 1 executing the program stored in the storage unit 12 and controlling the communication unit 13 and the like.

物品資訊取得部51，係將關於物品的複數個影像、和說明資訊，從使用者所操作的顧客終端2加以取得。複數個影像係可包含有：拍攝了該物品的影像、和不含該物品的說明用之影像。說明資訊係含有例如：像是廠商或產品名等這類表示物品之種類的資訊、購入時期、傷痕等之物品之狀態的說明文。The article information acquiring unit 51 acquires a plurality of images and explanatory information about the article from the customer terminal 2 operated by the user. A plurality of images may include: an image in which the item is photographed, and an image for explaining that does not include the item. The explanatory information is an explanatory text including, for example, information indicating the type of the article such as a manufacturer or a product name, the date of purchase, and the state of the article such as a scratch.

適合決定部52，係針對已被取得之複數個影像之每一者，決定對於物品之狀態之推定的適合性。物品之狀態之推定，係藉由狀態輸出部53中所含之影像推定部54及狀態算出部55而被進行。又，適合性係表示，能夠從影像多少程度地正確推定出物品之狀態。The suitability determination unit 52 decides suitability for estimation of the state of the article for each of the plurality of acquired images. The status of the article is estimated by the image estimation unit 54 and the status calculation unit 55 included in the status output unit 53 . Also, the suitability indicates how accurately the state of the article can be estimated from the video.

適合性之決定係亦可為，是否將該影像使用於藉由影像推定部54而被進行的獲得狀態之要素值的處理之判定，亦可為表示從該影像所被推定之值的信賴度的權重係數之算出。權重係數係為，藉由狀態算出部55而被進行的根據每一影像而被獲得的要素值來推定物品之狀態之際，表示該影像之重要度的值。The determination of suitability may be a determination of whether to use the image in the process of obtaining the element value of the state performed by the image estimation unit 54, or may indicate the reliability of the value estimated from the image. The calculation of the weight coefficient of . The weight coefficient is a value indicating the degree of importance of the image when the state of the article is estimated from the element value obtained for each image by the state calculation unit 55 .

狀態輸出部53，係基於適合決定部52之決定結果、和已被取得之複數個影像之至少一部分，而將表示該物品之已被推定之狀態的狀態資訊予以輸出。The state output unit 53 outputs state information indicating the estimated state of the article based on the determination result of the conformity determination unit 52 and at least a part of the plurality of acquired images.

狀態輸出部53中所含之影像推定部54，係針對複數個影像之至少一部分之每一者，算出物品之狀態是已被推定的要素值。The image estimation unit 54 included in the status output unit 53 calculates an estimated element value for each of at least a part of the plurality of images.

狀態輸出部53中所含之狀態算出部55，係基於針對複數個影像之每一者所被算出的要素值、與適合性，而將已被推定之物品之狀態進行推定，並輸出表示該已被推定之狀態的狀態資訊。The state calculation unit 55 included in the state output unit 53 estimates the state of the estimated article based on the element value and suitability calculated for each of a plurality of images, and outputs an output indicating the state of the article. Status information for presumed status.

圖3係為資訊處理伺服器1的處理之一例的說明用流程圖。首先，雖然未圖示，但顧客終端2係將用來登錄販售對象之物品的畫面予以輸出，將使用者於該畫面中所輸入的說明資訊、及已登錄之複數個影像，向資訊處理伺服器1進行送訊。FIG. 3 is an explanatory flowchart of an example of the processing of the information processing server 1 . First, although not shown in the figure, the customer terminal 2 outputs a screen for registering items to be sold, and sends the explanatory information input by the user on the screen and the registered images to the information processing unit. Server 1 sends the message.

物品資訊取得部51，係藉由顧客終端2，而將關於特定之物品的說明資訊與複數個影像，予以接收(步驟S101)。The article information acquisition unit 51 receives explanatory information and a plurality of images of a specific article through the customer terminal 2 (step S101).

圖4係為被輸入之影像之一例的圖示。在圖4的例子中，作為關於物品的影像，圖示了藉由使用者而被拍攝的物品也就是智慧型手機80之影像。在使用者所拍攝的影像中，與產品的廣告用之影像不同，經常會發生邊緣不鮮明的現象、或映入環境光之反射等。Fig. 4 is a diagram showing an example of an input image. In the example of FIG. 4, the image of the smartphone 80 which is the object captured by the user is shown as the image related to the object. In the image taken by the user, unlike the image used for product advertisement, blurred edges and reflections of ambient light often occur.

一旦接收複數個影像，則適合決定部52，係針對複數個影像之每一者，決定對於物品之狀態之推定的適合性(步驟S102)。一旦適合性被決定，則狀態輸出部53中所含之影像推定部54，係針對複數個影像之至少一部分之每一者，算出物品之狀態之要素值(步驟S103)。此處，在適合決定部52判定是否將複數個影像之每一者使用於影像推定部54之處理的情況下，則影像推定部54係可針對已被判定為要使用於處理的影像之每一者，算出要素值。又，在適合決定部52算出複數個影像之每一者的權重係數的情況下，則影像推定部54亦可針對所有的影像而算出要素值，亦可針對權重係數不是0的影像而算出要素值。Once a plurality of images are received, the suitability determination unit 52 determines suitability for estimation of the state of the article for each of the plurality of images (step S102 ). Once the suitability is determined, the image estimation unit 54 included in the state output unit 53 calculates the element value of the state of the article for each of at least a part of the plurality of images (step S103). Here, when the suitability determination unit 52 determines whether to use each of a plurality of images for the processing of the image estimation unit 54, the image estimation unit 54 may select each of the images determined to be used for processing. One, the element value is calculated. Also, when the suitability determination unit 52 calculates the weight coefficient for each of a plurality of images, the image estimation unit 54 may calculate element values for all images, or calculate element values for images whose weight coefficients are not 0. value.

要素值係可藉由機器學習模型而被算出。更具體而言，影像推定部54，係將複數個影像之每一者輸入至機器學習模型，將該機器學習模型之輸出當作從該影像所被推定出來的要素值而加以取得。Element values can be calculated by machine learning models. More specifically, the image estimation unit 54 inputs each of a plurality of images to a machine learning model, and obtains an output of the machine learning model as an element value estimated from the image.

機器學習模型，係預先藉由含有學習輸入影像與物品之狀態之判定結果的學習資料，而進行了學習。該學習資料中係含有：影像已被放大或縮小成所定像素數的學習輸入影像、和表示該影像之要素值的所給定之輸出資料。學習資料係亦可為，分別含有影像、與對該影像所被預先賦予的物品之狀態之要素值的複數個組合，亦可為分別含有從跳蚤市場服務所被抽出的物品之影像與對該物品所被賦予之排名(視為要素值)的複數個集合。The machine learning model is previously learned using learning data including learning input images and judgment results of the states of objects. The learning data includes a learning input image in which the image has been enlarged or reduced to a predetermined number of pixels, and given output data representing element values of the image. The learning material may be a plurality of combinations each including an image and an element value of the state of an item assigned to the image in advance, or may include an image of an item extracted from a flea market service and the value of an item associated with the image. Multiple sets of ranks (considered element values) assigned to items.

機器學習模型，在本實施形態中係可為：AdaBoost、隨機森林、神經網絡、支持向量機(SVM)、最近鄰識別器等之機器學習所被實作的機器學習模型。更具體而言，作為機器學習模型亦可使用所謂的Deep Learning來建構機器學習模型，在Deep Learning之中又亦可使用像是Attention Branch Network這類會一面自動學習在判定之際所注目之領域而一面學習要素值之推定的機器學習模型。In this embodiment, the machine learning model may be a machine learning model implemented by machine learning such as AdaBoost, random forest, neural network, support vector machine (SVM), and nearest neighbor recognizer. More specifically, as a machine learning model, so-called Deep Learning can also be used to construct a machine learning model. In Deep Learning, it is also possible to use such as Attention Branch Network, which automatically learns the area of attention when making a judgment. On the other hand, it is a machine learning model that learns the estimation of element values.

又，要素值，係亦可不將影像直接以機器學習模型進行處理就被算出。例如，亦可將藉由Bag of Visual Words(BoVW)之手法而從影像所被抽出之複數個Visual Word之每一者之頻率加以取得，基於該頻率而藉由所定之函數而算出要素值。此外，影像推定部54，係亦可使用藉由含有頻率之分布與物品之狀態的教師資料而被進行學習的機器學習模型，來判定特徵。Also, the element value may be calculated without directly processing the image with the machine learning model. For example, it is also possible to obtain the frequency of each of a plurality of Visual Words extracted from an image by the method of Bag of Visual Words (BoVW), and calculate an element value by a predetermined function based on the frequency. In addition, the image estimation unit 54 may also use a machine learning model learned from teacher data including the frequency distribution and the state of the item to determine the feature.

一旦要素值被算出，則狀態算出部55，係針對影像之至少一部分之每一者，基於物體之狀態之要素值和已被決定之適合性，來推定物品之狀態(步驟S104)。作為適合性係為判定是否將複數個影像之每一者使用於影像推定部54之處理的情況下，則可將已被推定之要素值之平均，當作物品之狀態之值而加以取得。Once the element values are calculated, the state calculating unit 55 estimates the state of the object based on the element values of the state of the object and the determined suitability for each of at least a part of the image (step S104). When suitability is to determine whether to use each of a plurality of images in the processing of the image estimation unit 54, the average of the estimated element values can be obtained as the value of the state of the article.

作為適合性係為算出權重係數的情況下則將要素值之加權平均當作物品之狀態而加以取得。更具體而言，狀態算出部55係將針對複數個影像之每一者而被算出的權重係數之合計，予以算出。狀態算出部55，係針對影像之每一者而算出要素值與權重係數的積，將該積之合計加以取得。狀態算出部55，係將積的合計除以權重係數之合計所得的值，當作表示物品之狀態的值而加以取得。In the case where the weight coefficient is calculated as suitability, the weighted average of element values is obtained as the state of the article. More specifically, the state calculation unit 55 calculates the total of the weight coefficients calculated for each of the plurality of images. The state calculation unit 55 calculates the product of the element value and the weight coefficient for each video, and acquires the total of the product. The state calculation unit 55 obtains a value obtained by dividing the total of the products by the total of the weight coefficients as a value representing the state of the article.

狀態算出部55，作為物品之狀態，亦可不是取得已被算出之物品之狀態之值本身，而是取得排名。此情況下，狀態算出部55係可為，排名與值之範圍是被預先建立對應，將已被算出之物品之狀態之值所屬之範圍所對應之排名，當作物品之狀態，亦即狀態資訊而加以取得。The state calculation unit 55 may acquire the rank instead of the value itself of the calculated state of the item as the state of the item. In this case, the state calculation unit 55 may be that the rank and the value range are associated in advance, and the rank corresponding to the range where the value of the calculated state of the item belongs is regarded as the state of the item, that is, the state information is obtained.

一旦物品之狀態被推定，則狀態算出部55係將該已被推定之物品之狀態予以輸出(步驟S105)。狀態算出部55，作為物品之狀態，亦可將令表示已被推定之物品之狀態的資訊被顧客終端2做顯示的資訊，予以輸出。此情況下，顧客終端2，係亦可將以使用者所輸出的狀態作為參考而輸入之表示物品之排名的資訊發送至資訊處理伺服器1，狀態算出部55係將其與該物品建立關連而記憶在記憶部12中。又，狀態算出部55，係亦可不經使用者之操作就向記憶部12輸出物品之狀態，令其與使用者所登錄之物品建立關連而將該物品之狀態加以記憶。 Once the state of the article is estimated, the state calculation unit 55 outputs the estimated state of the article (step S105). The state calculation unit 55 may output information indicating the estimated state of the article to be displayed on the customer terminal 2 as the state of the article. In this case, the customer terminal 2 may also send the information representing the ranking of the item input with the status output by the user as a reference to the information processing server 1, and the status calculation unit 55 may associate it with the item. And stored in the memory unit 12 . In addition, the state calculation unit 55 can also output the state of the item to the memory unit 12 without the user's operation, and make it associate with the item registered by the user to memorize the state of the item.

在本實施形態中，藉由影像推定部54(步驟S103)而針對影像之每一者來推定物品之狀態之要素值，然後藉由狀態算出部55(尤其是步驟S104)而根據要素值來推定物品本身之狀態。此處假設，從使用者所被登錄的複數個影像，是含有由廠商所作成的產品之影像或主要是含由說明文所成之影像的情況下，則從這些影像所被推定出來的狀態之要素值，恐怕會不正確。適合決定部52，係評估可從影像算出適切之要素值的或然性之高低，並反映在其以後之處理，藉此就可更正確地推定出物品之狀態。 In this embodiment, the element value of the state of the article is estimated for each of the images by the image estimation unit 54 (step S103), and then the state calculation unit 55 (especially step S104) calculates the value of the element from the element value. Presume the state of the item itself. Assuming here that the plurality of images registered by the user include images of products created by the manufacturer or mainly include images of explanatory texts, the state estimated from these images The value of the element may be incorrect. The suitability determination unit 52 evaluates the possibility of calculating the appropriate element value from the image and reflects it in the subsequent processing, so that the state of the article can be estimated more accurately.

進一步說明步驟S102之處理的細節。圖5係為適合決定部52的處理之一例的流程圖。圖5係圖示了，使用機器學習模型來決定適合性之情況的例子。 The details of the processing of step S102 are further described. FIG. 5 is a flowchart of an example of processing by the suitability determination unit 52 . Figure 5 illustrates an example of a situation where a machine learning model is used to determine suitability.

適合決定部52，係首先從已被取得之複數個影像，選擇出尚未被選擇的影像(步驟S201)。接著，適合決定部52，係使用機器學習模型，將表示已被選擇之影像 係為拍攝了特定之物品之影像的或然性之高低的適合性之值，加以取得(步驟S202)。 The fit determination unit 52 first selects an unselected image from the plurality of acquired images (step S201). Next, the fit determination unit 52 uses a machine learning model to represent the selected image It is acquired for the suitability value of the probability that the image of the specific object is photographed (step S202).

更具體而言，適合決定部52，係將已被選擇之影像輸入至機器學習模型，將該機器學習模型之輸出當作適合性而加以取得。機器學習模型，在本實施形態中係為例如：AdaBoost、隨機森林、神經網絡、支持向量機(SVM)、最近鄰識別器等之機器學習所被實作的機器學習模型。 More specifically, the suitability determination unit 52 inputs the selected image to the machine learning model, and acquires the output of the machine learning model as suitability. In this embodiment, the machine learning model is a machine learning model implemented by machine learning such as AdaBoost, random forest, neural network, support vector machine (SVM), and nearest neighbor recognizer.

機器學習模型，係預先藉由含有學習輸入影像與輸出值的訓練資料，而進行了學習。該學習資料中係含有：影像已被放大或縮小成所定像素數的學習輸入影像、和表示該影像之適合性的所給定之輸出值(例如權重係數之值)。 The machine learning model is pre-learned with training data including learning input images and output values. The learning material contains: a learning input image in which the image has been enlarged or reduced to a predetermined number of pixels, and a given output value (such as a weight coefficient value) indicating the suitability of the image.

一旦適合性被取得，則適合決定部52，係在有尚未被選擇之影像存在的情況下(步驟S203的Y)，重複步驟S201起之處理。另一方面，在所有影像都已被選擇的情況下(步驟S203的N)，就結束適合決定部52之處理。 Once the suitability is acquired, the suitability determination unit 52 repeats the processing from step S201 if there is an unselected video (Y in step S203). On the other hand, when all the images have been selected (N in step S203), the processing of the fit determination unit 52 ends.

於步驟S202中，亦可不必對機器學習模型輸入影像就決定適合性。例如，適合決定部52，係亦可使用Bag of Visual Words(BoVW)之手法而從影像來生成Visual Word之出現頻率之直方圖，藉由判定該直方圖之特徵而決定適合性。適合決定部52，係亦可使用藉由含有直方圖與適合性之值的教師資料而被進行學習的機器學習模型來判定特徵，亦可基於預先制定的算出演算法而從直方圖算 出適合性之值。 In step S202, the suitability may be determined without inputting images to the machine learning model. For example, the suitability determination unit 52 can also use the method of Bag of Visual Words (BoVW) to generate a histogram of the frequency of occurrence of Visual Words from the image, and determine the suitability by judging the characteristics of the histogram. The suitability determination unit 52 may also use a machine learning model learned from teacher data including histograms and suitability values to determine features, or may calculate features from the histogram based on a predetermined calculation algorithm. Get the value of suitability.

關於某個影像的適合性，係亦可藉由例如，偵測與該影像相同之影像是否存在於其他網站，而被決定。圖6係為適合決定部52的處理之另一例的流程圖。 The suitability of an image can also be determined by, for example, detecting whether images identical to the image exist on other websites. FIG. 6 is a flowchart of another example of the processing of the suitability determination unit 52 .

在圖6的例子中，首先，適合決定部52，係從複數個影像，將尚未被選擇之影像當作查詢影像而加以選擇(步驟S251)，將網域清單之中表示目前正被選擇之網域的資訊予以重置。適合決定部52，係從網域清單中，按照順序而選擇出1個網域(步驟S252)。網域清單，係為作為影像檢索對象之網域的清單。網域清單中所含之網域係可已被分群。網域中係被配置有複數個影像，已被選擇之網域中所被配置的影像，就成為檢索對象。網域係為用來特定出影像之提供來源的資訊。 In the example of FIG. 6, at first, the fit determination unit 52 selects an image that has not been selected as a query image from a plurality of images (step S251), and selects the image that is currently being selected in the domain list. Domain information is reset. The suitability determination unit 52 selects one network domain in order from the network domain list (step S252). The domain list is a list of domains to be searched for images. The domains contained in the domain list can be grouped. A plurality of images are arranged in the domain, and the images arranged in the selected domain become search objects. The domain is the information used to identify the source from which the image was provided.

圖7係為網域與權重係數之關係之一例的圖示。圖7所示的表，係表示了網域之群組與權重係數之關係。權重係數，係在群組中所屬之網域中存在有與查詢影像相同之影像的情況下，作為適合性而被決定。在圖7的例子中，第1個群組係含有商品的製造商及販售商之網域，第2個群組係含有讓中古品做流通的跳蚤市場或競標的網站之網域。 FIG. 7 is a diagram illustrating an example of the relationship between network domains and weight coefficients. The table shown in FIG. 7 shows the relationship between the group of the network domain and the weight coefficient. The weight coefficient is determined as suitability when the same image as the query image exists in the domain to which the group belongs. In the example shown in FIG. 7 , the first group contains domains of manufacturers and sellers of commodities, and the second group contains domains of flea markets or bidding websites that distribute second-hand goods.

例如商品的製造商及販售商之網域內，被配置有新品之廣告用之影像的或然性，係為極高。因此，該網域之群組所被建立關連的權重係數係設成0。另一方面，在無論哪個網域中都不存在相同之影像的情況下，則 其為自行拍攝之影像的或然性極高。於是，此情況的權重係數係設成最大的值也就是1。圖7的例子中作為權重係數之值，也會存在0與1之中間的值。例如，跳蚤市場或競標之網站的群組中所含之網域中，考慮相同物品會有被重複出品的可能性，而將權重係數之值設成中間的值。此外網域清單中亦可含有其他販售網站之網域。 For example, the probability that images for advertisements of new products will be placed in the domains of manufacturers and sellers of products is extremely high. Therefore, the weight coefficient of the association established in the group of the domain is set to 0. On the other hand, if the same image does not exist in either domain, then There is a high probability that it is an image taken by myself. Therefore, the weight coefficient in this case is set to the maximum value, which is 1. In the example of FIG. 7 , there may be a value between 0 and 1 as the value of the weight coefficient. For example, in the domains included in the group of websites for flea markets or bidding, consider the possibility that the same item may be repeatedly sold, and set the value of the weight coefficient to an intermediate value. In addition, the list of domains may also contain domains of other selling websites.

一旦網域被選擇，適合決定部52，係從含有已被選擇之網域的URL中所被配置的1或複數個影像，檢索出與查詢影像相同之影像(步驟S252)。適合決定部52係亦可使用，例如***影像檢索中所被使用的這類公知的網際網路上的影像檢索技術，從網域中所被配置之影像之中，檢索出與查詢影像相同之影像。又，在影像檢索中，亦可將類似影像之中類似度高於閾值者視為相同之影像來進行檢索。 Once the domain is selected, the suitability determination unit 52 retrieves the same image as the query image from one or more images arranged in the URL containing the selected domain (step S252). The suitability determination unit 52 may also use, for example, a known image retrieval technology on the Internet, such as is used in Google image retrieval, to retrieve an image identical to the query image from among the images arranged in the network domain. . In addition, in image retrieval, among similar images, those whose similarity degree is higher than a threshold value may be regarded as identical images and retrieved.

藉由檢索而有找到與查詢影像相同之影像的情況下(步驟S254的Y)，則適合決定部52，係將已被選擇之曾為檢索之對象的網域所被建立關連的權重係數，當作查詢影像之適合性而加以決定(步驟S255)。 In the case where an image identical to the query image is found through the search (Y in step S254), the appropriate determination unit 52 is the weight coefficient associated with the selected network domain that was once the object of the search, It is determined as the suitability of the query image (step S255).

另一方面，於步驟S254中，藉由檢索而未找到與查詢影像相同之影像的情況下(步驟S254的N)，則適合決定部52係判定是否有尚未被選擇之網域存在(步驟S256)。若有尚未被選擇之網域存在的情況下(步驟S256的Y)，則針對剩餘的網域，重複步驟S252以後之處理。若所有的網域都已經被選擇的情況下(步驟S256的N)，則適合決定部52係將最大的權重係數(在圖7的例子中係為1.0)，當作目前所被選擇之查詢影像之適合性而加以決定(步驟S257)。On the other hand, in step S254, when the image identical to the query image is not found by searching (N in step S254), the suitability determination unit 52 determines whether there is an unselected network domain (step S256 ). If there are unselected network domains (Y in step S256), the processing after step S252 is repeated for the remaining network domains. If all the network domains have been selected (N in step S256), the suitability determination unit 52 uses the largest weight coefficient (1.0 in the example of FIG. 7) as the currently selected query The suitability of the image is determined (step S257).

在步驟S255或步驟S257被執行後，適合決定部52係判定，是否有尚未被選擇作為查詢影像之影像存在(步驟S258)。若有尚未被選擇作為查詢影像之影像存在的情況下(步驟S258的Y)，則適合決定部52係重複步驟S251以後之處理，若所有的影像都已經被選擇作為查詢影像的情況下(步驟S258的N)，則結束處理。After step S255 or step S257 is executed, the suitability determination unit 52 determines whether there is an image that has not been selected as the query image (step S258 ). If there is an image that has not been selected as the query image (Y in step S258), then the suitability determination unit 52 repeats the processing after step S251, and if all the images have been selected as the query image (step S258). N) of S258, then end the processing.

此外，網域的清單係被配置成，權重係數越重的群組中所屬之網域，是在越前面的順序。假設彼此互異的權重係數所被建立關連的複數個網域中有相同之影像存在的情況下，則最低的權重係數係被當作查詢影像之適合性而被決定。In addition, the list of domains is configured such that the domains belonging to the group with the heavier weight factor are in the order of the front. Assuming that the same image exists in a plurality of domains associated with different weight coefficients, the lowest weight coefficient is determined as the suitability of the query image.

在圖6的例子中，隨著與已被使用者所登錄之影像相同之影像是否有從預先決定之網域被找到，以及，找到該影像之網域，來設定權重係數。亦可取而代之，適合決定部52係基於已找到的相同之影像之數量，來設定權重係數。例如，亦可為，找到的影像之數量超過第1閾值的情況下則將權重係數設成1，在其以下之情況下則隨應於所找到的影像之數量而藉由從0至1做單調增加的所定之函數，來計算權重係數。In the example of FIG. 6 , the weighting factor is set according to whether an image identical to an image registered by the user is found from a predetermined domain, and the domain where the image is found. Alternatively, the suitability determination unit 52 may set weighting coefficients based on the number of found identical images. For example, it is also possible to set the weight coefficient to 1 when the number of found images exceeds the first threshold, and to set the weight coefficient from 0 to 1 in accordance with the number of found images when it is below the first threshold. Monotonically increasing given function to calculate weight coefficients.

如目前為止所說明，像是中古商品這類對象為被特定之物品之狀態，是根據影像來做判定之際，決定將該影像使用於物品之狀態之推定的適合性，根據影像之每一者的適合性與從影像所被抽出的物品之狀態之要素值，來推定物品之狀態。藉此，即使使用者所登錄的複數個影像、且為關於物品的複數個影像，是未包含有該物品之影像的情況下，仍可更正確地推定該物品之狀態。又，也可根據已被推定之狀態而更正確地推測物品的價格(例如適切價格、建議價格、容易被購入的價格)，在跳蚤市場服務或競標服務中，可支援使用者設定金額的作業。As explained so far, when the state of a specified object such as a second-hand commodity is determined based on an image, the suitability of the image to be used for the estimation of the state of the object is determined based on each of the images. The suitability of the person and the element value of the state of the object extracted from the image are used to estimate the state of the object. Thereby, even if the plurality of images registered by the user are images about the item but do not include the item, the status of the item can be estimated more accurately. In addition, it is also possible to more accurately estimate the price of an item (such as a suitable price, a suggested price, and a price that is easy to buy) based on the estimated state, and it is possible to support the operation of setting the amount by the user in the flea market service or bidding service .

藉由使得被自動算出的物品之狀態變得更加正確，在讓使用者將中古品以CtoC進行販售的跳蚤市場服務或競標服務中，可以減少因為非專家的使用者來輸入物品之狀態所引起的物品之狀態之評價的偏誤。By making the automatically calculated item status more accurate, in the flea market service or bidding service that allows users to sell second-hand items as CtoC, it is possible to reduce the inconvenience caused by non-expert users inputting item status. The resulting bias in the evaluation of the state of the item.

目前為止所說明的技術係於跳蚤市場服務或競標服務中尤其有效，但只要是由使用者來登錄關於物品之複數個影像的服務，則都可適用。例如，亦可對中古商品的收購服務，適用本技術。此情況下，即使在已登錄的影像中，含有被印刷了商品之箱子的影像的這類情況下，仍可正確地偵測出物品之狀態。又，藉由將根據已被推定之狀態所預測的價格，當作暫定的收購預想金額而提示給使用者，也可支援使用者的賣出之判斷，也可減輕判定商品之狀態的事業者的負荷。The techniques described so far are particularly effective in flea market services and bidding services, but are applicable to any service in which a user registers a plurality of images of an item. For example, this technology can also be applied to the purchase service of second-hand goods. In this case, even if the image of the box on which the product is printed is included in the registered image, the state of the item can be detected correctly. In addition, by presenting the price predicted from the estimated state to the user as a tentative expected purchase price, it is also possible to support the user's judgment of selling, and it is also possible to reduce the burden on the business operator of judging the state of the product. load.

1:資訊處理伺服器 2:顧客終端 11:處理器 12:記憶部 13:通訊部 14:輸出入部 51:物品資訊取得部 52:適合決定部 53:狀態輸出部 54:影像推定部 55:狀態算出部 80:智慧型手機 1: Information processing server 2: Customer terminal 11: Processor 12: Memory Department 13: Department of Communications 14: I/O 51: Item information acquisition department 52: Fit Decision Department 53: Status output unit 54: Image Estimation Department 55:Status Calculation Department 80:Smartphone

[圖1]本發明的實施形態所述之資訊處理系統之一例的圖示。 [ Fig. 1 ] A diagram showing an example of an information processing system according to an embodiment of the present invention.

[圖2]資訊處理系統所實現之機能的區塊圖。 [Fig. 2] A block diagram of the functions realized by the information processing system.

[圖3]資訊處理伺服器的處理之一例的說明用流程圖。 [FIG. 3] A flow chart for explaining an example of the processing of the information processing server.

[圖4]所被輸入的影像之一例的圖示。 [FIG. 4] A diagram showing an example of an input image.

[圖5]適合決定部的處理之一例的流程圖。 [ Fig. 5 ] A flow chart of an example of the processing of the suitability determination unit.

[圖6]適合決定部的處理之另一例的流程圖。 [ Fig. 6 ] A flowchart of another example of the processing of the suitability determination unit.

[圖7]網域與權重係數之關係之一例的圖示。 [FIG. 7] A diagram showing an example of the relationship between domains and weight coefficients.

Claims (8)

一種資訊處理裝置，係含有：取得部，係取得關於特定之物品的複數個影像；和決定部，係針對前記複數個影像之每一者而決定對於前記物品之狀態之推定的適合性；和輸出部，係基於前記決定部之決定結果、與前記複數個影像之至少一部分，而將表示前記特定之物品的已被推定之狀態的狀態資訊，予以輸出；前記決定部，係以使得前記取得部所取得之影像係為拍攝了前記特定之物品之影像的或然性越高，該當影像的權重就會越大的方式，來決定權重。 An information processing device comprising: an acquisition unit that acquires a plurality of images related to a specific item; and a determination unit that determines, for each of the aforementioned plurality of images, the suitability of the presumption of the state of the aforementioned item; and The output unit outputs state information representing the estimated state of the item specified in the foregoing based on the determination result of the foregoing determination unit and at least a part of the plurality of images in the foregoing; The weight of the image obtained in the section is determined in such a way that the higher the probability that the image of the aforementioned specified item is captured, the greater the weight of the corresponding image. 如請求項1所記載之資訊處理裝置，其中，前記決定部，係將前記複數個影像之每一者的權重，加以決定；前記輸出部，係基於前記複數個影像之每一者，而將表示前記特定之物品之狀態的狀態要素值加以決定，並基於針對前記複數個影像之每一者而被決定的狀態要素值及前記權重，而生成表示前記特定之物品之狀態的狀態資訊，並將前記已被生成之狀態資訊予以輸出。 The information processing device as described in Claim 1, wherein the aforementioned determination unit determines the weight of each of the aforementioned plurality of images; the aforementioned output unit determines the weight of each of the aforementioned plurality of images. The state element value indicating the state of the item specified in the above is determined, and the state information indicating the state of the item specified in the above is generated based on the state element value determined for each of the plurality of images in the above and the weight in the above, and Output the status information that has been generated before. 如請求項1所記載之資訊處理裝置，其中，前記決定部係含有：藉由表示影像與前記影像之適合性的訓練資料而被進行學習的已學習模型。 In the information processing device as described in claim 1, wherein the antecedent determination unit includes: a learned model that is learned using training data indicating the suitability of the image and the antecedent image. 如請求項1所記載之資訊處理裝置，其中，前記決定部，係檢索與前記已被取得之複數個影像相同之影像，基於相同之影像是否藉由檢索而被找到，來決定前記適合性。 In the information processing device as described in claim 1, wherein the affidavit determining unit searches for images identical to the plurality of images obtained in the adacription, and determines the adequacy of the affidavit based on whether the same image is found through the search. 如請求項2所記載之資訊處理裝置，其中，前記決定部，係檢索與前記已被取得之複數個影像相同之影像，基於相同之影像是否藉由檢索而被找到，來決定前記權重。 In the information processing device as described in claim 2, wherein the affidavit determination unit searches for images identical to the plurality of images already obtained in the affidavit, and determines the affidavit weight based on whether the same image is found through the search. 如請求項5所記載之資訊處理裝置，其中，前記決定部，係在前記相同之影像是藉由檢索而有被找到的情況下，基於前記已被找到之影像的提供來源，來決定權重。 In the information processing device as described in claim 5, wherein the aforesaid determination unit determines the weight based on the source of the found image when the same aforesaid image is found by searching. 如請求項1所記載之資訊處理裝置，其中，前記決定部，係將前記取得部所取得之影像係為前記特定之物品所被拍攝之影像的或然性之高低，當作前記適合性而加以決定。 In the information processing device as described in Claim 1, wherein, the aforementioned determination unit determines the probability that the image obtained by the aforementioned acquisition unit is an image captured by the item specified in the aforementioned item as the suitability of the aforementioned item. to be decided. 一種資訊處理方法，係藉由電腦執行：取得關於特定之物品的複數個影像之步驟；和針對前記複數個影像之每一者而決定對於前記物品之狀態之推定的適合性之步驟；和 基於前記決定之結果、與前記複數個影像之至少一部分，而將表示前記特定之物品的已被推定之狀態的狀態資訊予以輸出之步驟；在前記進行決定的步驟中，係以使得前記進行取得的步驟所取得之影像係為拍攝了前記特定之物品之影像的或然性越高，該當影像的權重就會越大的方式，來決定權重。 An information processing method performed by a computer: a step of obtaining a plurality of images of a specific item; and a step of determining, for each of the aforementioned plurality of images, the suitability of the presumption of the state of the aforementioned item; and A step of outputting state information indicating the estimated state of the item specified in the foregoing based on the result of the determination in the foregoing and at least a part of a plurality of images in the foregoing; The weight is determined in such a way that the higher the probability that the image obtained in the step is that the image of the aforementioned specific item is captured, the greater the weight of the corresponding image will be.
TW110123302A 2020-10-29 2021-06-25 Information processing device and information processing method TWI793653B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2020181777A JP7092845B2 (en) 2020-10-29 2020-10-29 Information processing equipment and information processing method
JP2020-181777 2020-10-29

Publications (2)

Publication Number Publication Date
TW202217705A TW202217705A (en) 2022-05-01
TWI793653B true TWI793653B (en) 2023-02-21

Family

ID=81380295

Family Applications (1)

Application Number Title Priority Date Filing Date
TW110123302A TWI793653B (en) 2020-10-29 2021-06-25 Information processing device and information processing method

Country Status (3)

Country Link
US (1) US20220138926A1 (en)
JP (1) JP7092845B2 (en)
TW (1) TWI793653B (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040032979A1 (en) * 2002-08-12 2004-02-19 Hitachi High-Technologies Corporation Defect inspection method
CN103279871A (en) * 2013-06-06 2013-09-04 北京京东尚科信息技术有限公司 After-sales service application method and system
TW201504619A (en) * 2013-07-19 2015-02-01 Hon Hai Prec Ind Co Ltd System and method for detecting shape flaw
CN109559304A (en) * 2018-11-23 2019-04-02 聚时科技(上海)有限公司 Image quality online evaluation method, apparatus and application for industrial vision detection
CN110892349A (en) * 2017-05-23 2020-03-17 沃尔玛阿波罗有限责任公司 Automatic inspection system
US20200226731A1 (en) * 2019-01-15 2020-07-16 International Business Machines Corporation Product defect detection

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9721292B2 (en) * 2012-12-21 2017-08-01 Ebay Inc. System and method for image quality scoring

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040032979A1 (en) * 2002-08-12 2004-02-19 Hitachi High-Technologies Corporation Defect inspection method
CN103279871A (en) * 2013-06-06 2013-09-04 北京京东尚科信息技术有限公司 After-sales service application method and system
TW201504619A (en) * 2013-07-19 2015-02-01 Hon Hai Prec Ind Co Ltd System and method for detecting shape flaw
CN110892349A (en) * 2017-05-23 2020-03-17 沃尔玛阿波罗有限责任公司 Automatic inspection system
CN109559304A (en) * 2018-11-23 2019-04-02 聚时科技(上海)有限公司 Image quality online evaluation method, apparatus and application for industrial vision detection
US20200226731A1 (en) * 2019-01-15 2020-07-16 International Business Machines Corporation Product defect detection

Also Published As

Publication number Publication date
US20220138926A1 (en) 2022-05-05
JP2022072384A (en) 2022-05-17
TW202217705A (en) 2022-05-01
JP7092845B2 (en) 2022-06-28

Similar Documents

Publication Publication Date Title
CN110378731B (en) Method, device, server and storage medium for acquiring user portrait
US11816721B2 (en) Recommendation apparatus and method
US20230394432A1 (en) Machine-learned model for optmizing selection sequence for items in a warehouse
US9082086B2 (en) Adaptively learning a similarity model
JP2012529122A (en) System and method for learning user genre and style and matching products to user preferences
JP7162417B2 (en) Estimation device, estimation method, and estimation program
US20120296776A1 (en) Adaptive interactive search
US20190108571A1 (en) System and method for efficiently determining relevant products for display in an online marketplace
US20190287154A1 (en) Image searching apparatus, printed material, and image searching method
JP2023052120A (en) Customer personalization control unit, system, and method
US20160171365A1 (en) Consumer preferences forecasting and trends finding
JP2018045505A (en) Determination device, determination method, and determination program
CN112424812A (en) Calculating user-specific item-related values on an electronic processing platform
TWI793653B (en) Information processing device and information processing method
US20170039620A1 (en) Server, analysis method and computer program product for analyzing recognition information and combination information
US20160307220A1 (en) Non-transitory computer readable medium, information processing apparatus, and information processing method
CN111639274B (en) Online commodity intelligent sorting method, device, computer equipment and storage medium
US20210065276A1 (en) Information processing apparatus and non-transitory computer readable medium
CN111127128B (en) Commodity recommendation method, commodity recommendation device and storage medium
JP6958014B2 (en) Recommender systems, information processing devices and programs
JP7127749B2 (en) Recommended information identifying device, recommended information identifying system, recommended information identifying method, and program
JP7127728B2 (en) Recommended information identifying device, recommended information identifying system, recommended information identifying method, terminal device, and program
JP7326364B2 (en) Information processing device, information processing method and information processing program
JP7115655B2 (en) Recommended information identifying device, recommended information identifying system, recommended information identifying method, and program
CN110377831B (en) Retrieval method, retrieval device, readable storage medium and electronic equipment